Anchoring Bias in Large Language Models: An Experimental Study
It addresses a cognitive bias problem in LLMs for AI safety and fairness, but it is incremental as it focuses on a specific bias with limited mitigation solutions.
This study investigated anchoring bias in large language models (LLMs) like GPT-4 and Gemini, finding that LLM responses are sensitive to biased hints and that simple mitigation strategies such as Chain-of-Thought are insufficient, requiring comprehensive hint collection instead.
Large Language Models (LLMs) like GPT-4 and Gemini have significantly advanced artificial intelligence by enabling machines to generate and comprehend human-like text. Despite their impressive capabilities, LLMs are not immune to limitations, including various biases. While much research has explored demographic biases, the cognitive biases in LLMs have not been equally scrutinized. This study delves into anchoring bias, a cognitive bias where initial information disproportionately influences judgment. Utilizing an experimental dataset, we examine how anchoring bias manifests in LLMs and verify the effectiveness of various mitigation strategies. Our findings highlight the sensitivity of LLM responses to biased hints. At the same time, our experiments show that, to mitigate anchoring bias, one needs to collect hints from comprehensive angles to prevent the LLMs from being anchored to individual pieces of information, while simple algorithms such as Chain-of-Thought, Thoughts of Principles, Ignoring Anchor Hints, and Reflection are not sufficient.